In this paper, we present anew approach to automatic alignment of multi-word terms based on contextual information. With the English-Chinese Parallel Corpus of Maritime Conventions (ECPCMC) as a basis, we first extract the candidate terms from the corpus of each language separately using a hybrid approach which combines a list of user-adjustable morphosyntactic patterns and a term weighting measure. Then the extracted terms are aligned based on the contextual information with the assumption that translation equivalents are often used in similar contexts. For this purpose, we extract the collocates of each candidate term in either side of the languages in the corpus and comparethem for the degree of similarity at the semantic attribute level. Since the collocation extraction is restricted to the concordance lines containing the candidate term, this method not only overcomes the limitations of conventional statistical methods requiring large corpora to be effective but also improves the accuracy of automatic alignment. Experiments based on a section of ECPCMC, which is annotated with terms manually, indicate that the proposed approach is promising, giving an encouraging 85.3% precision and 60.6% recall on average.